Free models

What "free" means here

Every entry below carries a free quota — an allowance the provider grants, usually on signup or as a promotion. It is not a free tier of ours, and it is not unlimited. Three shapes show up most often:

What the provider grants	How it behaves
Token allowance with an expiry, typically 1M tokens over 90 days	The common case. Spends down as you use it, disappears on the date.
Outright free	No expiry announced. Providers change this without notice.
Rate-limited plan: no token cap, just a ceiling on how often you can call	No token cap, but a ceiling on how fast. Fine for interactive use, bad for batch.

Free quota belongs to the provider. We cannot extend it, restore it, or promise it will still be there next month — which is exactly why the router checks the balance at request time rather than trusting a list.

That quota — and the rate limit that comes with it — is the provider's; we set no limit of our own. Our own free tier is separate: fm-v1-lite, no subscription, no charge, and it never silently turns into a paid model. Paid tiers (fm-v1-standard, fm-v1-pro) are reached only when you name them and hold a subscription.

How to check right now

The model list is public. No key needed.

# the models callable right now
curl -s https://freemodel.online/v1/models

The list is live. If a model is missing from it, it is not callable right now — this page is a copy of what the endpoint says today.

Choosing a tier instead of a model

FreeModel publishes three tier aliases. Each one expands to a pool of models rather than naming a single one, and the pools differ in what they cost you:

Model id	Behaviour
`fm-v1-lite`	The free tier. Any account can call it, with no subscription.
`fm-v1-standard`	Needs a standard monthly plan. Without one, the call is refused with a 402.
`fm-v1-pro`	Needs a pro monthly plan. Without one, the call is refused with a 402.

fm-v1-lite is the only free tier, and it is the one to send when a surprise bill would be worse than a failed request. The other two refuse rather than bill: without the matching monthly plan the request comes back with a 402.

Finding free capacity for a task

The routes above are the whole mechanism. In practice, this is the order the decision runs in.

Step 1: Start from the free tier. Send fm-v1-lite. It is the free tier, callable by any account without a subscription, so it is the one to pick when a charge is what you are avoiding.
Step 2: Narrow the task before you narrow the route. The modality is the tightest filter: only 153 of the 199 listed models are chat models, so a request that generates an image or transcribes speech starts from a much smaller pool than a chat request does.
Step 3: Expect to fall through on the rarer modalities. Chat is 77% of the list, so a route into image, video or speech has far fewer entries to draw free quota from.[1]
Step 4: Read the log entry if a request fails. Every skipped candidate carries its reason; when all of them say the same thing, the problem is the request rather than the route.
Step 5: Move up a tier once you want the stronger pools. fm-v1-standard and fm-v1-pro draw from higher-scoring models, and each needs its monthly plan; without it the call comes back with a 402 rather than a bill.

Why this page prints no number

The absence of a count is deliberate and worth explaining. FreeModel does not know how much an allowance has left in it until it asks the provider, and the public list carries a modality rather than a balance. An allowance is a value the provider holds and changes without telling anyone: it is consumed by use, refilled on a schedule, and withdrawn when a promotion ends.

77% of the 199 listed entries are chat models and 9% generate images;[1] the allowance question cuts across the modalities rather than sitting inside one of them. So it is answered per request by the router, not per entry by a list somebody keeps updating. That is also why the route syntax exists: auto/image expands to the image candidates at request time, not a lookup in a table compiled last week.

What an allowance does not cover

It does not cover every modality. Chat is 77% of the list; image, video and speech are the rest.[1] A request outside chat is the one most likely to leave the allowance behind and land on a listed rate.
It does not last. Grants expire on the provider's schedule. Most of them expire unused precisely because that schedule is nobody's calendar but the provider's.
It does not transfer. An allowance belongs to one provider and one model family. Nothing converts the remainder at one vendor into capacity at another.
It is not ours to extend. We cannot restore a spent grant, move its expiry date, or promise that it will still exist next month. That is why the balance is read when the request is made rather than trusted from a list.

Free quota FAQ

How do I know whether a quota is left before I send a request?

From outside, you cannot — the public list does not expose it. The router reads the remaining balance at request time, which is more accurate than any number this page could print, and picks accordingly.

Why did a request I expected to be free cost money?

Because a grant is a preference, not a filter. The router prefers entries with quota left while the allowance lasts and bills the listed rate after it. To keep a request off that rate, send fm-v1-lite: it is the free tier, any account can call it, and it needs no subscription.

Does spent quota ever come back?

Sometimes. Providers refill on their own schedule, and new promotions appear without notice. Nothing in FreeModel controls either of those events, and no setting in the console changes them.

Can I top up a provider's grant through FreeModel?

No. A grant belongs to the provider that issued it. What the console sells is your own balance, which is spent only after the available free capacity has been used.

Is "free quota" the same as a free tier?

No. There is no free tier of ours. Some providers hand out an allowance, the router prefers those entries while the allowance lasts, and everything after that is billed at the listed per-model rate.

Sources

Every third-party figure quoted on this site was read from the vendor documentation linked below, on the date stated. None of these links are affiliate links and none of these vendors paid to be listed. If a number here disagrees with the page it came from, the page is right — tell us and it gets corrected.

Where the percentages come from

Every share quoted on this site is counted from the live model list, not estimated. Divide the two counts in the last column, round to the nearest whole percent, and you get the number in the first.

Share	Of what	Count
77%	are chat models	153 of 199 listed entries
9%	generate images	18 of 199 listed entries
23%	are not chat models	46 of 199 listed entries
3%	are modality routes	6 of 199 listed entries
17%	carry a context length	34 of 199 listed entries

[1] Counted on 29 September 2026 from GET https://freemodel.online/v1/models, which needs no key and no account. Every row uses all 199 listed entries as the denominator, because the only thing the endpoint states for every entry is its modality. It no longer publishes per-model capability flags such as tool calling or reasoning, so this site quotes no shares for them. Fetch the list and recount whenever you want to check the arithmetic.

Models that come with quota you did not pay for